1 00:00:00,566 --> 00:00:02,366 if AI is not citing your content 2 00:00:02,366 --> 00:00:04,566 the problem may not be your expertise 3 00:00:04,566 --> 00:00:07,133 it may be your file format 4 00:00:07,133 --> 00:00:07,400 so 5 00:00:07,400 --> 00:00:11,166 a lot of companies are publishing solid content that humans can read 6 00:00:11,166 --> 00:00:14,700 just fine but AI systems can't reliably fetch it 7 00:00:14,700 --> 00:00:16,033 structure it ground it 8 00:00:16,033 --> 00:00:18,000 or tie it back to one clear source 9 00:00:18,766 --> 00:00:21,866 when that happens the model either skips you or cites someone else 10 00:00:21,866 --> 00:00:24,666 or fills in the blanks so in this video 11 00:00:24,666 --> 00:00:26,300 I wanna break down which 12 00:00:26,466 --> 00:00:30,566 type of files that LLMs actually prefer for citations 13 00:00:30,566 --> 00:00:31,433 and more importantly 14 00:00:31,433 --> 00:00:35,100 how to package your content so AI can trust it fast hi 15 00:00:35,100 --> 00:00:38,300 I'm Kevin Roy of Green Banana SEO and search is changing fast 16 00:00:38,300 --> 00:00:40,766 and people aren't clicking the 10 blue links first 17 00:00:40,833 --> 00:00:42,433 they're getting an answer from AI 18 00:00:42,433 --> 00:00:44,166 so the new goal isn't just to rank 19 00:00:44,166 --> 00:00:46,566 it's to become the source that AI sites 20 00:00:46,666 --> 00:00:49,633 and on this channel I'm gonna show you the exact no BS 21 00:00:49,633 --> 00:00:52,233 no buzzword marketing junk that strings you along 22 00:00:52,233 --> 00:00:53,900 wait for an answer that never comes 23 00:00:54,066 --> 00:00:57,466 I'm gonna show you the systems that we use here to earn visibility 24 00:00:57,466 --> 00:01:00,466 in Google AI overviews AI mode chat GPT 25 00:01:00,466 --> 00:01:01,500 Gemini Perplexity 26 00:01:01,500 --> 00:01:03,466 and more we cover content architecture 27 00:01:03,466 --> 00:01:05,433 empty signals schema measurement 28 00:01:05,433 --> 00:01:06,733 and anything else that works 29 00:01:06,966 --> 00:01:08,866 so these videos are built as a framework 30 00:01:08,866 --> 00:01:11,266 so that you can copy and leave with a plan 31 00:01:11,266 --> 00:01:12,566 not just ideas 32 00:01:13,433 --> 00:01:15,100 now here's the big idea 33 00:01:16,033 --> 00:01:19,433 llms don't just prefer one magical file type 34 00:01:19,433 --> 00:01:23,066 they prefer clean citation package 35 00:01:23,200 --> 00:01:25,800 and that package usually looks like this 36 00:01:25,800 --> 00:01:31,366 a clean HTML page a readable JSON version of the same content 37 00:01:31,500 --> 00:01:36,800 a durable PDF mirror and if you have video or audio or a transcript 38 00:01:37,166 --> 00:01:39,200 in the right media schema 39 00:01:39,200 --> 00:01:42,500 that's the real play that's why I'm doing these videos 40 00:01:43,366 --> 00:01:45,600 let's start with the most important move 41 00:01:45,600 --> 00:01:47,666 for every important piece of content 42 00:01:47,666 --> 00:01:50,400 publish a machine readable JSON import 43 00:01:50,466 --> 00:01:55,466 that could be something like a claim dot Json or a post dot Json 44 00:01:55,466 --> 00:01:56,800 why does that matter 45 00:01:56,833 --> 00:02:01,633 because now you're not forcing AI systems to guess what the page is 46 00:02:01,633 --> 00:02:04,833 you're handing them a structured version with a stable ID 47 00:02:04,833 --> 00:02:06,833 a source URL a title 48 00:02:06,833 --> 00:02:08,666 a summary and a publish date 49 00:02:08,766 --> 00:02:11,066 that makes grounding for them easier 50 00:02:11,066 --> 00:02:12,366 then on the page itself 51 00:02:12,366 --> 00:02:15,966 your visible headline should clearly match the claim in the JSON 52 00:02:15,966 --> 00:02:18,066 so this JSON is schema by the way 53 00:02:18,866 --> 00:02:20,800 it's a huge mistake that people make 54 00:02:20,866 --> 00:02:22,666 their page headline says one thing 55 00:02:22,666 --> 00:02:24,766 and their meta data says something else 56 00:02:24,766 --> 00:02:27,300 and then the body wanders off into three different ideas 57 00:02:27,300 --> 00:02:30,900 that creates ambiguity and ambiguity kills citations 58 00:02:31,200 --> 00:02:32,866 so making the page obvious 59 00:02:32,866 --> 00:02:35,566 one strong headline one clear claim 60 00:02:35,566 --> 00:02:38,900 one source of truth pdfs 61 00:02:39,033 --> 00:02:40,633 a lot of people overlook pdfs 62 00:02:40,633 --> 00:02:42,033 I personally don't like them 63 00:02:42,033 --> 00:02:44,766 but they are still very strong supporting an asset 64 00:02:44,766 --> 00:02:46,300 because they tend to be durable 65 00:02:46,300 --> 00:02:48,600 portable and consistently indexed 66 00:02:48,900 --> 00:02:50,600 so if you have an important article 67 00:02:50,600 --> 00:02:52,466 a report or a research page 68 00:02:52,466 --> 00:02:57,166 create a clean PDF mirror of it and don't just export it and forget it 69 00:02:57,200 --> 00:02:58,266 add a strong XMP meta data title 70 00:03:00,066 --> 00:03:01,266 author published date 71 00:03:01,266 --> 00:03:03,233 description and canonical URL 72 00:03:03,233 --> 00:03:05,866 and then you can host it at a predictable URL 73 00:03:05,866 --> 00:03:07,300 and link it from a main page 74 00:03:07,300 --> 00:03:09,633 you can put all that stuff inside of a PDF 75 00:03:09,633 --> 00:03:14,233 and why because you have another stable version of the same content 76 00:03:14,233 --> 00:03:15,966 that points back to the source 77 00:03:16,266 --> 00:03:18,033 now let's talk multimedia 78 00:03:18,033 --> 00:03:20,566 if you publish video or podcast content 79 00:03:20,566 --> 00:03:23,366 and you want AI systems to ground answers to it 80 00:03:23,366 --> 00:03:27,000 don't stop at the embed you need a transcript 81 00:03:27,000 --> 00:03:33,400 ideally a dot vtt or a dot SRT file as a stable URL 82 00:03:33,400 --> 00:03:38,333 then add a video object audio object schema with the name description 83 00:03:38,400 --> 00:03:41,200 duration content URL and bed URL 84 00:03:41,200 --> 00:03:44,700 update date and transcript if appropriate 85 00:03:45,566 --> 00:03:49,000 this is one of the easiest wins right now because without a transcript 86 00:03:49,000 --> 00:03:52,500 you're asking the model to infer meaning from the media 87 00:03:52,500 --> 00:03:53,400 with a transcript 88 00:03:53,400 --> 00:03:56,400 you're actually giving it the text to retrieve and compare it 89 00:03:56,400 --> 00:03:59,933 and quote now if you publish research studies 90 00:04:00,033 --> 00:04:02,733 original data this matters even more 91 00:04:03,066 --> 00:04:06,400 those pages should have a dataset level structure 92 00:04:06,400 --> 00:04:08,700 that means a simple dataset 93 00:04:08,700 --> 00:04:11,666 .Json file versioning distribution links 94 00:04:11,666 --> 00:04:17,633 a CSV and daily checksums and source references if you have them why 95 00:04:17,633 --> 00:04:20,800 because AI systems trust content more 96 00:04:20,800 --> 00:04:24,766 when it's easy to verify and easier to tie to a stable source 97 00:04:25,133 --> 00:04:27,166 so if I had to simplify this whole thing 98 00:04:27,166 --> 00:04:29,000 into one sentence it would be this 99 00:04:29,000 --> 00:04:31,166 llms don't prefer one file type 100 00:04:31,166 --> 00:04:34,733 they prefer consistency across formats 101 00:04:34,966 --> 00:04:38,333 clean HTML matching Json Schema 102 00:04:38,366 --> 00:04:41,433 durable Pdfs transcript back media 103 00:04:41,433 --> 00:04:43,966 stable URLs clean metadata 104 00:04:43,966 --> 00:04:46,966 and that's how you make content citation ready 105 00:04:47,000 --> 00:04:48,200 and before you publish it 106 00:04:48,200 --> 00:04:51,100 just run a fast QA check 107 00:04:51,633 --> 00:04:56,166 does the on page claim match exactly the Json of your schema claim 108 00:04:56,233 --> 00:04:59,066 can the Json slash schema be fetched cleanly 109 00:04:59,066 --> 00:05:02,366 does the PDF work point back to the canonical URL 110 00:05:02,566 --> 00:05:04,366 basically points back to the link 111 00:05:05,766 --> 00:05:09,600 if it's a video do the transcripts and schema resolve or they match 112 00:05:09,600 --> 00:05:12,066 if it's these pieces line up you're giving AI 113 00:05:12,066 --> 00:05:15,433 systems a much better shot at grounding answers in your content 114 00:05:15,433 --> 00:05:16,466 instead of guessing 115 00:05:17,000 --> 00:05:18,566 that's the real takeaway here 116 00:05:18,566 --> 00:05:19,100 THANK YOU